摘要
针对现场烧结矿FeO控制复杂与难度大的实际,采用改进后的4层前向神经网络,进行多因素输入建模,输出采用主因线性相关与次因非线性相关叠加,预报烧结矿FeO,为现场烧结矿FeO控制提供了新的可行方法。该网络结构设计先进合理,精度高,泛化能力强。训练误差平方和为0.0794,用训练样本集测试FeO输出,检验的绝对平均误差为0.109467,命中率97.81%。采用训练后网络预报,其绝对平均误差为0.1068255,命中率100%。
Aim at actual complicacy and difficulty of controlling FeO content in sinter ,the model of improved 4 layers feedforward neural network has been set up, its input is multi-factor manner and its output is constituted by adding of the main factor linear relativity to hypo-factor nonlinear relativity . This model can provide a new method for FeO content control in production field. The network possess advanced reasonable construction designs, high accuracy and strong generalization ability. The network training sum of squared error is 0.0794. To test the output of FeO with the training sample set, the absolute average error of examination is 0.109467, the rate in accuracy area of checkout is 97.81%. By using forecasts after network training, its absolute average error is 0.1068255, the rate in accuracy area of forecasts is 100%.
出处
《烧结球团》
北大核心
2005年第3期30-34,共5页
Sintering and Pelletizing
关键词
人工神经网络
烧结矿
FEO
网络训练
预报系统
sinter FeO content, influence factors, neural network improving design, modeling, network training, forecast
